[ros-users] [Discourse.ros.org] [Jobs] About the Jobs category

Stan Boland ros.discourse at gmail.com
Tue Mar 28 18:41:38 UTC 2017

**Research Scientists  Prediction, Behaviour and Motion Planning**
**Cambridge and Edinburgh, UK**

Five AI is a venture-backed start-up based in the UK that is developing a full autonomous vehicle software stack for driverless vehicles.  Its intention is to build a system that is safe in complex urban environments and that requires us to research, develop, integrate and test the worlds leading state-of-the-art SLAM, object recognition, scene understanding, 3D reconstruction, behaviour recognition, scene prediction, motion planning and vehicle control technologies. Its business model is to deliver that technology in a service model in our towns and cities.  The Company was founded in late 2015 by a talented group of algorithm, software and semiconductor specialists with a strong track record of building successful companies and was seed funded in summer 2016 by some of Europes strongest early stage venture investors.
We are now seeking three highly talented Research Scientists focused on prediction, action, behaviour and motion planning as a key competency.

**Job Description**
	Research and then develop state-of-the-art large scale predictive models of dynamic road scene environments to produce rich predictions, dynamic cost maps and motion plans 
	Explore and develop richly-structured end-to-end deep neural networks to provide intermediation between perception inputs and useful outputs, such as dynamic costmaps
	Apply novel prediction techniques such as inverse reinforcement learning (inverse optimal control) to generate those cost maps; find novel ways of using vast quantities of un-annotated video footage to learn those cost functions
	Develop and improve ways of establishing beliefs in the goals and behaviours of actors in road scenes and develop and curate learnt behavioural model libraries
	Research and identify suitable methods of building memory into these models, for example using LSTMs and RNNs
	Research novel means of natively modelling interactive and dynamic motions using those beliefs, dynamic costmaps and kinematic models of each actor, such as dynamic programming-based solvers
	Explore and implement game theoretic approaches alongside reinforcement learning to predict interactive actor behaviour in road scenes and their likelihoods
	Find ways of using deep perception outputs to improve those predictions (gestures, pose, gaze, wheel movements etc) and identify means of feeding those requirements back to the perception layers of the run-time vehicle stack 
	Select and develop path planning methodologies, likely based on variants of D*, A*, RRT* algorithms, which, together with kino-dynamic vehicle models, deliver feasible and preferred paths
	Generate novel research techniques for solving multiple challenges in the field of computer vision as applied to autonomous vehicles and work to get outputs published in the worlds top scientific journals and conferences
Education and Experience
	PhD graduate in the field of machine learning and/or artificial intelligence under supervision of some of the worlds leading robotics and/or machine learning research leaders
	Masters-level graduate in a relevant field (computer science, mathematics, physics, statistics, engineering) from a recognised leading school or university in the field 
	At least one year of post-doctoral research and/or industry experience
	Track record of work being published in top scientific journals and/or being accepted at leading international conferences (eg IROS, AAAI, IJRR, NIPS, ICRA, MLJ, JMLR, HSR, EUMAS etc)

**Core Skills**
	Strong background in machine learning as applied to prediction, action, behaviour, autonomous and lifelong agent learning, Bayesian methods in decision making etc
	Strong analytical background and deep understanding of large-scale optimization, numerical linear algebra, probabilistic inference
	Strong programming skills (eg Python, C/C++, Scala, Matlab, Java) as well as LINUX, HTML and GPU programming including CUDA
	Excellent communication skills (fluency in English essential)
	Desire to work in a fast-paced start-up environment

	Salary at least 45,000, depending on experience
	Early stock options in venture-backed start-up
	Health insurance and life assurance cover

	Applications should be made through Five AIs website www.five.ai
	Closing date for applications is 30 April 2017

[Visit Topic](https://discourse.ros.org/t/about-the-jobs-category/301/2) or reply to this email to respond.

More information about the ros-users mailing list